Trustworthy Artificial Intelligence

Your challenge

The use of algorithms, including those that use AI techniques (hereafter together referred to as ‘algorithms’), has increased greatly over the last years. Algorithms make it possible to execute tasks more efficiently and make more effective and data driven decisions. However, as the use of algorithms comes with opportunities, it also comes with specific risks and requirements. When these risks are not properly identified and managed, they can have harmful consequences to users, clients and society as a whole. Therefore, it is important that safeguards are in place across the lifecycle of the algorithm. These safeguards ensure that results of algorithms are trusted and that the algorithms are used in an ethical and responsible way. The responsible use of algorithms is crucial if we are making decisions for, or about, clients and society as a whole.

Insights

Het beheer van algoritmerisico’s met een algoritmekader

Het gebruik van algoritmes is de afgelopen jaren sterk toegenomen binnen overheidsinstellingen en andere publieke instellingen. Algoritmes kunnen zorgen voor meer efficiëntie, effectiviteit en snelheid in besluitvorming en de uitvoering van beleid. Daarentegen brengen algoritmes ook nieuwe risico’s met zich mee. 

Lees hier meer over 'Het beheer van algoritmerisico’s met een algoritmekader'

Een algoritme hoeft geen risico te zijn

Het gebruik van algoritmes binnen de overheid is in de afgelopen jaren sterk toegenomen. De besluitvorming aan de hand van het algoritme kan discriminerend zijn, moeilijk uitlegbaar of kan de privacy van burgers schenden. Dit artikel gaat verder in op een aantal aanbevelingen van de Algemene Rekenkamer om deze risico's te verminderen.

Lees hier meer over 'Een algoritme hoeft geen risico te zijn'

Our Trustworthy AI Framework

We put trust at the centre of everything we do. The multidimensional trustworthy AI framework helps organisations to develop safeguards for algorithms across six key dimensions. These safeguards are a crucial step in managing risks whilst capitalising on the returns associated with algorithms.

Deloitte can help you to ensure your algorithms and organisation as a whole adhere to these six dimensions of the trustworthy AI framework.

Our solutions and credentials

Deloitte has experience with addressing a variety of challenges that come with AI within your organisation, including:

• AI risk and quality management:
  We help organisations with the design, implementation and operation of AI risk and quality management frameworks. Accountability is ensured by clearly defining the right governance, roles and responsibilities, policies and documents, and algorithm registries throughout the entire algorithm lifecycle.
• Independent algorithm review
  We help organisations with executing an independent algorithm review. We analyse whether an individual algorithm has a responsible business case, is fit-for-purpose and is in line with regulatory requirements and ethical values. Our approach is based, among others, on frameworks from the Algemene Rekenkamer and the IAMA. Our approach is tailored to the context of the algorithm.
• Organisational and regulatory readiness (e.g., accountability, policies, inventory and registry, risk appetite, monitoring and reporting)
  We help organisations to use algorithms responsibly. Our approach often includes an assessment of the maturity and ambition of the AI management capabilities of an organisation, including the readiness to adapt to future regulations such as the EU AI Act.
• Fairness & Bias
  We help organisations with implementing fairness safeguards across the algorithm lifecycle. We developed a bias detection and mitigation framework that helps organisations to identify and (if necessary) mitigate undesirable bias in data or an algorithm.
• Transparency & explainability
  We help organisations defining what type of transparency and explainability requirements are necessary for the different stakeholders throughout the algorithm lifecycle, and how to effectively communicate to stakeholders and society.
• Robustness
  We help organisations to achieve, validate and monitor the performance of their algorithms, focusing on reliability, stability, bias, assumptions.
• Privacy
  We help organisations to implement privacy principles and adhere to privacy regulations such as the GDPR during the development and use of algorithms.
• Safe & secure
  We help organisations to ensure that data and algorithms are used in a safe and secure environment and that IT controls such as backups are in place.
• Generative AI: 
  We help organisations to identify risks and formulating guidelines to combine Trustworthy and Generative AI using Large Language Models (LLMs) such as ChatGPT.

Why Deloitte

Trustworthy AI is at the core of responsible business by focusing on responsible, safe, fair and explainable use of algorithms.

Deloitte has experience with clients across various industries such as the financial, governance and public, consumer and technology, media and telecom sectors. Each sector and client have their own challenges. Hence, the solutions Deloitte provide are tailored to your sector and organisation.

Our focus on responsible business and experience with AI risk and quality management frameworks, set us apart. Our experience with developing and implementing these frameworks over the past years, enables us to bring this experience to you.

Please feel free to contact our Trustworthy AI team to discover how Deloitte can help your organization with a responsible use and implementation of artificial intelligence.